1.a.-Supercomputers.md

 There are couple supercomputers available to BSC. 
 Each have their own hardware and software stack and need to be managed differently.
-This page will give an overview of the basic information, but we *strongly* encourage you to view the official documentation of the supercomputers. All supercomputers use the [Slurm Workload Manager [1.1]](https://slurm.schedmd.com/documentation.html) for managing jobs, which we will not cover. 
+This page will give an overview of the basic information, but we *strongly* encourage you to view the official documentation of the supercomputers. All supercomputers use the [[a.1] Slurm Workload Manager](https://slurm.schedmd.com/documentation.html) for managing jobs, which we will not cover. 
 
 ## MareNostrum 4
-*[official documentation [1.2]](https://bsc.es/supportkc/docs/MareNostrum4/intro/)*  
+*[[a.2] official documentation](https://bsc.es/supportkc/docs/MareNostrum4/intro/)*  
 MareNostrum 4 is located at BSC and offers only a CPU partition.
 Hence, GPU offloading is not an option.
 However, it could function nicely as a different platform to test sequential and CPU optimized versions of your code.
 
 ## Lumi
-*[official documentation[1.3]](https://docs.lumi-supercomputer.eu/)*  
+*[[a.3] official documentation](https://docs.lumi-supercomputer.eu/)*  
 The Lumi supercomputer is divided in two partitions: [Lumi-C](https://docs.lumi-supercomputer.eu/hardware/lumic/) containing CPU nodes and [Lumi-G](https://docs.lumi-supercomputer.eu/hardware/lumig/) containing GPU nodes.
 The internal hardware structure of these partitions is different, and thus a comparison between the two is mostly of no use. 
 In other words, the CPU version of your code should be executed on Lumi-G as well when comparing to the GPU version.
@@ -87,7 +87,7 @@ The list below contains modules that are required when building specific project
 Be careful with experimenting with loading different module combinations. We noticed that some modules overwrite loaded architectures from previously loaded modules, which breaks their inner workings and causes errors. 
 
 ## MeluXina
-*[official documentation [1.4]](https://docs.lxp.lu/)*  
+*[[a.4] official documentation](https://docs.lxp.lu/)*  
 The MeluXina supercomputer consists of modules for CPU, GPU, FPGA, and Large memory applications. 
 The internal hardware structure of these partitions is different, and thus a comparison between the two is mostly of no use. 
 In other words, the CPU version of your code should be executed on the GPU module as well when comparing to the GPU version.
@@ -140,7 +140,7 @@ We generally make a distinction between 3 use-cases:
 **nemo-build:** There are no extra dependencies. HDF5 and NETCDF are build manually to avoid conflicts.
 
 ## Sources
-- [1.1 SLURM workload manager](0.-Sources#11-slurm-workload-manager)
-- [1.2 MareNostrum4 documentation](0.-Sources#12-marenostrum4-documentation)
-- [1.3 Lumi documentation](0.-Sources#13-lumi-documentation)
-- [1.4 MeluXina documentation](0.-Sources#14-meluxina-documentation)
+- [a.1 SLURM workload manager](0.a.-Documentation-&-Manuals#a1-slurm-workload-manager)
+- [a.2 MareNostrum4 documentation](0.a.-Documentation-&-Manuals#a2-marenostrum4-documentation)
+- [a.3 Lumi documentation](0.a.-Documentation-&-Manuals#a3-lumi-documentation)
+- [a.4 MeluXina documentation](0.a.-Documentation-&-Manuals#a4-meluxina-documentation)
\ No newline at end of file